BP网络用于枣香味成分的定量结构色谱保留关系研究  

Study on the QSRR of Aroma Compounds in Jujube by Back-Propagation Artificial Neural Network

作　　者：黄保军[1] 

机构地区：[1]河南省微纳米能量储存与转换材料重点实验室/许昌学院表面微纳米材料研究所,河南许昌461000

出　　处：《湖北农业科学》2013年第8期1927-1930,共4页Hubei Agricultural Sciences

基　　金：河南省科技厅国际科技合作资助项目(124300510054);河南省教育厅自然科学资助项目(2010B150026);许昌市科技计划项目(5007);许昌学院校内科研基金项目(2013067)

摘　　要：采用误差反传前向人工神经网络(BP-ANN)建立了63种大枣香味成分的结构与色谱保留之间的定量关系模型(QSRR)。以63种大枣香味成分的分子电性距离矢量为输入参数,色谱保留时间为输出参数,采用内外双重验证法分析该模型的稳定性和外部预测能力。所构建网络模型的相关系数为0.998 9,交叉检验相关系数为0.998 9,标准偏差为0.959,残差绝对值低于3.844,应用于外部预测集,外部预测集相关系数为0.998 9;而多元线性回归(MLR)法模型的相关系数为0.981 9,交叉检验相关系数为0.982 0,标准偏差为3.697、残差绝对值低于9.264,外部预测集相关系数为0.986 1。结果表明,ANN模型的拟合效果明显优于MLR模型。The systematic study of the quantitative structure-retention relationship（QSRR） on 63 aroma compounds in jujube was performed by the model of artificial neural network（ANN） based on the back propagation algorithm.The two-factor authentication method was carried out to analyze and test the stability and external predictive power of the models,when using the molecular electronegativity-distance vector（MEDV） of 63 aroma compounds in jujube as the inputs of the neural network and the retention times of these compounds as the outputs of the neural network.The correlation coefficient of the model was 0.998 9,the leave one out cross-validation regression coefficient was 0.998 9,the standard deviation was 0.959,the absolute value of residual was less than 3.844 and the correlation coefficient of the test set was 0.998 9.In order to make contrast,the QSRR model was set up by multiple linear regressions（MLR） method.For the model built by MLR,the correlation coefficient was 0.981 9,the leave one out cross-validation regression coefficient was 0.982 0,the standard deviation was 3.697,the absolute values of residual were less than 9.264 and the correlation coefficient of the test set was 0.986 1.The results showed that the performance of neural network method was better than that of MLR method.

关 键 词：大枣 香气成分 定量结构色谱保留关系(QSRR) 人工神经网络(ANN) 

分 类 号：TP29[自动化与计算机技术—检测技术与自动化装置] O657.7[自动化与计算机技术—控制科学与工程]